Adverse Consequences of Adverse Selection
Mark V. Pauly, Sean Nicholson
Journal of Health Politics, Policy and Law, Volume 24, Number 5, October
1999, pp. 921-930 (Article)
Published by Duke University Press
For additional information about this article
https://muse.jhu.edu/article/15418
Accessed 28 Jul 2017 17:18 GMT
JHPPL 24.5-04.Pauly
8/2/00
12:15 PM
Page 921
Adverse Consequences of
Adverse Selection
Mark Pauly and Sean Nicholson
University of Pennsylvania
The turnabout in public opinion about managed care has been especially striking. In the space of only a few years, managed care went
from being the centerpiece of many proposals for health reform to
being the target of legislative action that forbids it the managerial
actions that constitute its essence. What happened? Were the initial members of managed care plans overly accepting of restrictions they later
discovered to be oppressive? Did plans change over time to become
more aggressive? Or, is the controversy only political? We wish to suggest that there was more to the backlash than political mobilization of
a few inevitable malcontents; rather, neither initial buyers nor original
plans changed their views or behavior. Instead, a crucial part of the
change, and a potential cause of the backlash, was a change in the kinds
of people who were joining managed care plans.1 The changes we will
highlight are different from the kinds that can be easily brought into the
policy analytical framework because they represent transfers among
different sets of insured persons. The policy process must address the
much more vexing question of who should gain and who should lose,
with the total approximately constant. The source of this difficult problem, as with so many other problems in health insurance, is adverse
selection.
1. We do not argue that this is the only source of backlash. Consumer misinformation and tax
incentives to employers to choose insurance for their employees also contribute (Pauly and
Berger 1998). Here we highlight a neglected reason.
Journal of Health Politics, Policy and Law, Vol. 24, No. 5, October 1999. Copyright © 1999 by
Duke University Press.
JHPPL 24.5-04.Pauly
922
8/2/00
12:15 PM
Page 922
Journal of Health Politics, Policy and Law
Adverse Selection in Insurance: Theory
There is an economic theory of adverse selection in health insurance,
well known among economists in general, that so far has had few realworld applications. This theory was developed by Michael Rothschild
and Joseph Stiglitz (1976) to explain how competitive insurance markets
would behave in a world in which everyone knows that buyers differ by
risk levels but insurers are unable to distinguish among risks. Rothschild
and Stiglitz show that, if potential insurance purchasers know their risk
levels (if they can keep that information from insurers) and if insurers
are willing to offer any potentially profitable contracts, a process of sorting or self-selection can ensue.
This process works in stages. Assume that the market begins with a single policy offered. If insurers cannot distinguish well among risk levels,
the premium for the policy will be approximately “community rated”;
that is, it will be similar for all potential buyers. However, such a situation cannot be an equilibrium if insurers are free to enter and if consumers
know which risk type they are. Since high-risk consumers value generous
benefits more than do low-risk consumers, there will always exist another
policy, less generous in effective coverage than the community-rated one,
but lower in premium that will cause different risk levels to separate
themselves. In the case of two risk classes, high and low, there will always
be a policy that the low risks will prefer to the community-rated policy,
while the high risks will prefer to stick with the old policy at the old premium. In the next stage, once that new policy is offered (and the low risks
drain off), the old policy can no longer be profitably offered at its original
premium; since it is now chosen only by high risks, its premium will have
to rise. After the premium has increased, it is possible that the high risks
will now prefer the new policy, with less generous coverage to the old
community-rated policy; the original policy will disappear and everyone
will move in yet another stage to the new, limited policy. As we will note
below, however, even this third stage is not the end of the story. In fact, the
Rothschild-Stiglitz model displays the intriguing if upsetting property that
there may be no equilibrium; the market may move cyclically from one
policy to another, never settling down.
Adverse Selection in Insurance:
Application to Managed Care
Does this model help to explain what has been happening so far in many
managed care markets? The first, or community-rated, stage seems to
JHPPL 24.5-04.Pauly
8/2/00
12:15 PM
Page 923
Pauly and Nicholson
■
Adverse Selection
923
correspond roughly to what insurance markets were like before managed
care started to grow. Policies took the form of indemnity insurance and
were similar in terms of coverage. Insurance offered on the open market
to small- and medium-sized firms was only lightly experience rated.
(Self-insurance practiced by large groups is obviously perfectly based on
group experience.) Next, there is good evidence that oftentimes (though
not necessarily always) managed care plans offered alongside the old
comprehensive indemnity or “Blue Cross–type” plan tended to attract
people who were low risks (Jackson-Beeck and Kleinman 1983; Luft
1981). The requirement of changing physicians, and the expectation of
consumers that managed care coverage was better for people with mild
illnesses who wished to avoid hospitals and heroic measures (as compared to people with more serious illnesses or those who preferred more
aggressive treatment), often made managed care enrollments nonrandom, attracting disproportionate numbers of those who had been low utilizers in the indemnity plan.
These features would correspond to the first and second stages of the
Rothschild-Stiglitz model, in which an initial policy that was purchased
by both high and low risks at approximately pooled premiums set the
stage for the low risks to be attracted to a new policy with less generous coverage but lower cost. Of course, analysts have always been wary
of catastrophic indemnity insurance as an insurance attractive to low
risks, whereas managed care coverage, with only minimal cost sharing,
appeared to be more generous coverage than a moderate deductible Blue
Cross policy. However, there is more to coverage than the extent of outof-pocket payment. Coverage is also defined by the restrictiveness of use
of care and the type of provider; on this score, managed care is definitely
less generous.
It is the next stage of the Rothschild-Stiglitz model that we think is relevant to the managed care backlash. As lower risks were attracted away
from conventional insurance, its breakeven premiums had to rise. Once
that happened, there was yet another stage in which many of the higher
risks decided to migrate to managed care. Our key insight is that the
process of making this move literally was, for the higher risks, only making the best of a bad lot. They would have much preferred their initial situation with moderate premiums and permissive indemnity coverage, but
the selection process destroyed that option; their only choice now was
between a much more expensive indemnity plan and a lower-cost but
more constraining managed care plan. No wonder that the high-risk consumers became upset, especially as it dawned on them more and more
that managed care was not the kind of coverage they had been used to.
JHPPL 24.5-04.Pauly
924
8/2/00
12:15 PM
Page 924
Journal of Health Politics, Policy and Law
There is yet another stage of the Rothschild-Stiglitz theory to be mentioned because it seems to describe what is happening right now in the
world of transitions we are describing. Once the premium for the generous plan rose to the higher level and most of the market (including moderately high risks) had been pushed into more restrictive coverage, it
would no longer be necessary for that type of insurance to be so restrictive. If there were few remaining high risks for restrictiveness to keep
out, there was less need for limits on networks or access. So, in a paradoxical but understandable development, managed care recently became
less restrictive in many markets, even as the volume of complaints about
its restrictiveness grew. At the same time, indemnity insurance became
more “managed.” Networks expanded to cover the great majority of
providers in a market, pretreatment approval requirements were relaxed,
and external processes for settling disputes between insured and managed care plans were put in place by the plan. In effect, the market was
returning to something very close to the original pooled plan. (Indeed,
the out-of-network, point-of-service coverage now so popular in managed care plans does have cost-sharing provisions often almost identical
to the old Blue Cross plans.)
There is, as befits a model with cycles, yet another stage of the Rothschild-Stiglitz model and one which may describe the health insurance
market yet to come. But let us postpone that consideration in order to
review the empirical evidence in support of the interpretation we have
postulated.
Empirical Evidence in Support of the Model
There is some convincing empirical evidence supporting the RothschildStiglitz model. Figure 1 presents data on the percentage of employees at
large firms who were enrolled in indemnity and managed care plans in
1984 through 1998. It seems reasonable to characterize the health insurance market in 1984 as a pooling equilibrium. Over 90 percent of employees had indemnity insurance, mostly with Blue Cross and Blue Shield,
which experience rated only reluctantly. This pooling equilibrium unraveled between 1984 and the early 1990s when health maintenance organizations (HMOs) and preferred provider organizations (PPOs) quadrupled their share of the large employer market. By 1998, the original
market shares had been almost exactly reversed; managed care plans collectively had an 86 percent share of the large-employer market and
indemnity plans appeared to be in the final throes of a death spiral.
JHPPL 24.5-04.Pauly
8/2/00
12:15 PM
Page 925
Pauly and Nicholson
■
Adverse Selection
925
Figure 1 Enrollment by Type of Insurance Plan at Firms with 200+
Employees. Source: KPMG surveys of employer-sponsored health benefits, 1984 –1998.
In the previous section we argue that HMOs initially picked off lowrisk individuals who were least likely to object to restrictions on utilization of services and physician choice. If the difference between indemnity
and managed care premiums had remained constant over time, presumably the average risk of each insurance type would have remained constant. Figure 2 displays the average annual percentage change in indemnity and HMO premiums between 1981 and 1998. Although the rates of
change in premiums generally move in tandem, indemnity premiums grew
substantially more than HMO premiums between 1986 and 1991. This
pattern would be expected to occur as low risks shifted into managed care.
As the opportunity cost of remaining in an indemnity plan increased,
higher-risk people eventually switched, perhaps with apprehension, to
managed care plans; after 1991 the growth rates became similar.
We argued above that once a sufficient number of high risks join managed care plans, the plans (and low-risk enrollees) would become less
interested in maintaining the restrictions that were intended to keep the
high-risk individuals out in the first place. There is strong support for
this hypothesis. Information on the size of HMO provider networks is
reported in Table 1. Between 1990 and 1997, the average number of hospitals per HMO plan doubled, and the average number of primary care
physicians and specialists per HMO plan nearly tripled. Consumers are
JHPPL 24.5-04.Pauly
926
8/2/00
12:15 PM
Page 926
Journal of Health Politics, Policy and Law
Figure 2 Average Annual Percentage Change in Health Insurance
Premiums. Source: KPMG and A. Foster Higgins and Co., Inc., survey
of employer-sponsored health plans, 1991–1998; Health Insurance
Association of America surveys.
clearly attracted to plans that combine broad choice with a low premium
(relative to indemnity). Managed care plans that cover some portion of
out-of-network care (PPO and POS plans) have increased their enrollment among large employers during the last two years while non-POS
HMO market share has actually fallen (Figure 1).
While managed care plans were beginning to resemble indemnity
plans by increasing the size of their provider panels, indemnity plans
were adopting managed care techniques in an attempt to control costs,
offer competitive premiums, and maintain market share. In 1993, 52 percent of the indemnity plans were requiring precertification for elective
surgery (reported at the bottom of Figure 2). By 1996, almost threequarters of indemnity plans were requiring such precertification. Perhaps
we have come full circle; the health insurance market in 1998 could be
characterized, once again, as an approximate pooling equilibrium (or at
least an equilibrium with pooling of unobservable risk differentials). In
the early 1980s people of different risk levels were enrolled, for the most
part, in a similar type of indemnity plan; now people of different risk levels are enrolled, for the most part, in a similar lightly managed care plan.
There do not appear to be big differences between types of managed care
JHPPL 24.5-04.Pauly
8/2/00
12:15 PM
Page 927
Pauly and Nicholson
Table 1
■
Adverse Selection
927
Mean Number of Providers per HMO Plan
Primary Care
Physicians/Plan
Specialists/Plan
Hospitals/Plan
1990
1992
1995
1997
342
446
820
982
762
18
856
19
1,876
33
2,203
37
Source: Managed Care Digest, Hoechst Marion Roussel, 1991 – 1998.
plans. In 1997, for example, the average number of hospital days per thousand PPO and HMO enrollees was 252 and 231, respectively. Although
a nonnegligible percentage of employees at large firms are still covered
by an indemnity plan (14 percent), even these individuals are likely to
have their care managed to some extent.
Implications of the Rothschild-Stiglitz
Model for Citizen Dissatisfaction
If high- and low-risk individuals are currently pooled in managed care
plans, the Rothschild-Stiglitz model suggests that they might both be dissatisfied. High-risk individuals who dislike utilization management and
restrictions on physician choice might resent the restrictions associated
with managed care and may be unhappy about indemnity premium
increases. If these individuals joined a managed care plan recently, they
will not appreciate that managed care plans have actually become less
restrictive over time; their comparison will be to the indemnity plan they
left because its premiums became unacceptably high. Low-risk individuals might also be dissatisfied with managed care because they are
increasingly subsidizing high-risk individuals. As long as managed care
premiums continue to rise slowly, however, the low-risk individuals will
not realize the extent to which they are subsidizing the high risks.
Over thirty thousand individuals in sixty different markets were surveyed in 1996 as part of the Robert Wood Johnson Community Tracking
Study. Table 2 presents data from this attitudinal survey of the HMO
respondents toward health providers, their insurance company, and
health care generally. Survey respondents were asked a series of twelve
questions regarding their physical health and were assigned a score
according to the SF-12 methodology (Ware, Kosinski, and Keller 1995).
Individuals who were enrolled in an HMO at the time of the survey are
assigned in Table 2 to one of three groups depending on whether their
JHPPL 24.5-04.Pauly
928
8/2/00
12:15 PM
Page 928
Journal of Health Politics, Policy and Law
Table 2
HMO Enrollees’ Attitudes about Health Care
Organized by Category of Physical Health (SF-12)
Bottom Quartile
(Least Healthy)
n = 2,649
Very satisfied with health
care
Very satisfied with choice
of primary care physician
Very satisfied with choice
of specialist
Thoroughness of last exam/
treatment was very good or
excellent
Strongly agree that physician
may not refer respondent
to specialist when needed
Strongly agree that physician
is strongly influenced by
insurance company rules
Middle Two
Quartiles
n = 5,069
Top Quartile
(Healthiest)
n = 2,629
0.540
0.611*
0.625
0.600
0.637*
0.642*
0.658
0.713*
0.725*
0.678
0.715*
0.740*
0.107
0.063*
0.059*
0.264
0.210*
0.225*
Source: Community Tracking Study household survey.
*Significantly different, at the 5 percent level, from the mean value for the least healthy people.
SF-12 score is in the lowest quartile among this population, the middle
two quartiles, or the top quartile. High-risk individuals are clearly less
satisfied with their health plan. A relatively small percentage of individuals in the worst physical health are very satisfied with health care generally, their choice of primary care physicians and specialists, and the
quality of the care they receive. High-risk individuals are also more
likely to agree that their physician is influenced by HMO incentives and
that their physician restricts access to specialists.
What Should Happen Next?
If this is an acceptable explanation of how the backlash came to be, what
are the appropriate reactions from those concerned about public policy?
Most people might be upset that the previous cross subsidization of high
risks by low risks was disrupted. However, it is important to note that this
disruption was only temporary. Now that the traditional Blue Cross plan
has virtually disappeared as an option for the middle class, the different
JHPPL 24.5-04.Pauly
8/2/00
12:15 PM
Page 929
Pauly and Nicholson
■
Adverse Selection
929
risk levels again find themselves sharing a common insurance plan. A
major difference is that other low risks and high risks do not like it as
much as their old (but infeasible) plans. High risks find it more restrictive, and lower risks find it less of a bargain than their original, low-risk
HMO. Despite our argument about the moderation of managed care
aggressiveness, there are some differences in insurer behavior which are
likely to persist forever with no turning back to the good old days unless
regulators require it.
What are the messages for regulatory policy? Despite the bad feelings
and the nostalgia, we suspect that few would want to go back to the old
Blue Cross–type policy entirely. The notions of emphasis on preventive
care (even when it is not so cost effective), and the idea that some insurer
oversight of physician behavior is in the interest of consumer satisfaction, are probably here to stay.
What is more important in our view is for public policy to deal with the
adverse selection cycle. One more insight from the Rothschild-Stiglitz
model is that the somewhat more generous pooled policy that we currently seem to be moving toward in many markets will not be a permanent destination. Instead, once we arrive there, the original process of
self-selection will start all over again. The Rothschild-Stiglitz model tells
us what sort of thing to look for — some coverage provision that will
appeal to the low risks but not pick off the higher risks. Low payments to
providers leading to a “self-shrunken” network already appear to be one
such device, though there will doubtless be others.
There is, however, an alternative to having policy makers try to guess
what slick insurers (and greedy low risks) will come up with next as a
cherry-picking device. Everyone could be made permanently better off,
and the cycle avoided, if risk levels could be measured and adjusted for.
That is, some type of risk adjustment could help low risks to retain the
preventive-care-encouraging, inpatient-care-discouraging features of
group- or staff-model managed care plans.
How such adjustments would work in private markets is far from clear.
Employers or purchasing groups might decide to risk adjust the employer
payment. Regulators might enforce transfers through high-risk pools or
forbid coverage provisions targeted at lower risks. Tax policy makers
might adjust the tax subsidy to be more generous to high risks. But at
least we know one of the directions in which we should look.
JHPPL 24.5-04.Pauly
930
8/2/00
12:15 PM
Page 930
Journal of Health Politics, Policy and Law
References
Jackson-Beeck, Marilyn, and Joel Kleinman. 1983. Evidence for Self-Selection
a mong Health Maintenance Organization Enrollees. Journal of the American
Medical Association 250(November):2826 – 2829.
Luft, Harold S. 1981. Health Maintenance Organizations: Dimensions of Performance. New York: Wiley.
Pauly, Mark V., and Marc L. Berger. 1998. Why Should Managed Care and Managed
Care Insurance Be Regulated? Paper presented at the Fifth Princeton Conference,
March.
Rothschild, Michael, and Joseph Stiglitz. 1976. Equilibrium in Competitive Insurance
Markets: An Essay on the Economics of Imperfect Information. Quarterly Journal
of Economics 90(November):630 – 649.
Ware, John E., Mark Kosinski, and Susan D. Keller. 1995. SF-12: How the Score the
SF-12 Physical and Mental Health Summary Scales. 2d ed. Boston: The Health
Institute, New England Medical Center.